Laser-induced Autofluorescence Diagnosis Of Colorectal Cancer Optical System Design And Spectral Signal Processing

The dissertation is focused on the laser design reconstruction of autofluorescent detection system for colorectal cancer and the digital signal processing of collected autofluorescent spectra to determine optimal excitation wavelength and acquire characteristic indexes for discrimination cancerous from normal colorectal tissue. The research in this dissertation is funded by the national key technological project of the Ninth-Five-Project as the early colon autofluorescent endoscopy diagnosis system. Our study will help to build up diagnostic modal and develop autofluorescence detection system for clinical applications.The research demonstrates that Nd:YAG and Ti diamond Multiple output solid laser is effective in application of colorectal cancerous autofluorescence detection. We add Ti diomand parameter amplifier into laser optical path design, which provides hardware support for the determination of optimal wavelength. Based on the analysis of experimental data ,we determine the optimal wavelength is at 380 nm.As to the overwhelming large number of spectra data, without programming the digital signal processing is not available ,so multiple analysis of autofluorescent spectra data on the basis of Matlab software programming is necessary and effective. The signal data ofnormal and cancerous colorectal tissue are analyzed by slope-method, perceptron, mean value and variance method, and power spectra method respectively. The results from slope method obtain pretty good sensitivity and accordance rate, while results from power spectra are little lower. Mean value and variance method describes the spectral distribution properties of normal and cancerous tissues quantitatively. So taking the slope method as the main body, synthesis of the indexes of power spectra and mean value-variance will benefit for clinical diagnosis.Multiple analysis of autofluorescent spectra data on the basis of Matlab software programming is unique domestically ,and adopting the mean value-variance and power spectral method originated from the random signal processing to autofluorescent spectra analysis is another innovation in our dissertation.